The NAND (Negated AND or NOT AND) media check, the utility checks for bad blocks in the NAND flash memory that is used for the internal boot media. If found, the utility can recover the bad block. On very rare occasions, the file system cannot be recovered even after the recovery of bad blocks by the NAND media check utility. This article describes a recovery method of file system corruption using single user mode.

Example1:
Wrong information is found on a mounted partition, bo0s3f in this example.

root@SRX% fsck /dev/bo0s3f
** /dev/bo0s3f (NO WRITE)
** Last Mounted on /cf/var
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
FREE BLK COUNT(S) WRONG IN SUPERBLK
SALVAGE? no

SUMMARY INFORMATION BAD
SALVAGE? no

BLK(S) MISSING IN BIT MAPS
SALVAGE? no

529 files, 77049 used, 98275 free (539 frags, 12217 blocks, 0.3% fragmentation)

Example2:
The following is another case showing that single user mode is helpful.

Note:   In addition to the example described above, this article is helpful for a Junos OS upgrade when file system corruption is observed.

File system corruption might be caused mainly but not limited to the following:

• Ungraceful power off; customer operation, system crash etc.

• High frequent data writing to the flash; traffic log etc.

Regardless what the cause is, the solution below can help to recover the NAND flash to keep the device working normally.

1. Perform the bad block recovery using nand-mediack command (no -C option).

Note : At this point, the NAND media check utility checks the corrupted blocks and those corrupted blocks will be recovered.

2. Perform fsck -f on mounted partitions, and fsck -f -y on non-mounted partitions.

There are four partitions in flash media; Slice1(s1a), Slice2(s2a), and Slice3(s3e, s3f).  It will perform a "NO-WRITE" operation on mounted partitions, and a "WRITE" operation on non-mounted partitions.  Mounted and non-mounted partitions can be checked using the df command.

Example: If da0s1a is the primary root partition, it will perform a "NO-WRITE" operation on da0s1a and a "WRITE" operation on da0s2a .  bo0s3e and bo0s3f are mounted partitions, so a "NO-WRITE" operation will be performed.

root@% df
Filesystem   512-blocks   Used  Avail Capacity  Mounted on
/dev/da0s1a     1248744 301376 847472    26%    /  devfs                 2      2      0   100%    /dev
/dev/md0         797400 797400      0   100%    /junos
/cf             1248744 301376 847472    26%    /junos/cf
devfs                 2      2      0   100%    /junos/dev/
procfs                8      8      0   100%    /proc
/dev/bo0s3e       94304     60  86700     0%    /config  /dev/bo0s3f     1264808 956712 206912    82%    /cf/var  /dev/md1         687744  36456 596272     6%    /mfs
/cf/var/jail    1264808 956712 206912    82%    /jail/var
/cf/var/log     1264808 956712 206912    82%    /jail/var/log
devfs                 2      2      0   100%    /jail/dev
/dev/md2         128728      8 118424     0%    /mfs/var/run/utm
/dev/md3           3768      8   3460     0%    /jail/mfs

Mounted Partition
root@% fsck -f /dev/da0s1a
root@% fsck -f /dev/bo0s3e
root@% fsck -f /dev/bo0s3f

Non-Mounted Partition
root@% fsck -f -y /dev/da0s2a

At this point, file system recovery will be recovered for non-mounted partition by "WRITE" operation.

3. If file system error is observed on mounted partition, it is necessary to log in to single user mode during next reboot.

1. When the autoboot is completed, press the spacebar a few times to access the bootstrap loader prompt.
2. At the following prompt, enter boot -s to start up the system in single-user mode.
3. Enter full pathname of shell or 'recovery' for root password recovery or RETURN for /bin/sh: [Press Enter/Return, do not edit 'recovery']

1. Reboot the unit, and log in to Single User Mode (see above) during the bootup.

2. Check the mounted filesystem from Single User Mode.

# df
Filesystem  512-blocks   Used  Avail Capacity  Mounted on
/dev/da0s1a     598104 300084 250172    55%    /  
devfs                2      2      0   100%    /dev
/dev/md0        795116 795116      0   100%    /junos
/cf             598104 300084 250172    55%    /junos/cf
devfs                2      2      0   100%    /junos/dev/  

3. Perform fsck -f -y on non-mounted partitions ( bo0s3e,bo0s3f ).

Non-Mounted Partitions
root@% fsck -f -y /dev/bo0s3e
root@% fsck -f -y /dev/bo0s3f

4. Boot the unit.

• Install the upgrade image.

• Reboot to upgrade the unit.

1. Install the upgrade image => Primary root partition will be switched (e.g. da0s2a ).

2. Reboot the unit, and log in to Single User Mode (see above) during the bootup.

3. Check the mounted filesystem.

# df
Filesystem  512-blocks   Used  Avail Capacity  Mounted on
/dev/da0s2a    1264552 288416 874972    25%    /  
devfs                2      2      0   100%    /dev
/dev/md0        760316 760316      0   100%    /junos
/cf            1264552 288416 874972    25%    /junos/cf
devfs                2      2      0   100%    /junos/dev/

4. Perform fsck -f y on former primary root partition.

Non-Mounted Partition
root@% fsck -f -y /dev/da0s1a

5. Boot the unit.